Leveraging Large Language Models for Enhanced Classification and Analysis: Fire Incidents Case Study

Author:
Open Access
Publisher website
Google Scholar
Abstract
Fire detection and analysis have been a central focus of numerous studies due to their importance in potentially reducing fire’s harmful impact. Fire detection and classification using artificial intelligence (AI) methods have drawn significant attention in the literature. These methods often tackle certain aspects of fire, such as classifying fire versus non-fire images or detecting smoke or flames. However, these studies lack emphasis on integrating the capabilities of large language models for fire classification. This study explores the potential of large language models, especially ChatGPT-4, in fire classification tasks. In particular, we utilize ChatGPT-4 for the first time to develop a classification approach for fire incidents. We evaluate this approach using two benchmark datasets: the Forest Fire dataset and the DFAN dataset. The results indicate that ChatGPT has significant potential for timely fire classification, making it a promising tool to complement existing fire detection technologies. Furthermore, it has the capability to provide users with more thorough information about the type of burning objects and risk level. By integrating ChatGPT, detection systems can benefit from the rapid analysis capabilities of ChatGPT to enhance response times and improve accuracy. Additionally, its ability to provide context-rich information can support better decision-making during fire episodes, making the system more effective overall. The study also examines the limitations of using ChatGPT for classification tasks.

This publication has 32 references indexed in Scilit: